Constructions Of Prediction Models Of Brown Plant Hopper Occurrence Degree Based On Artificial Neural Network

The short-term, medium-term A and medium-term B prediction models of the brown plant hopper, Nilaparvata lugens (Stal), occurrence degree were constructed based on the artificial neural network by the Back Propagation Arithmetic. The data of brown plant hopper from 1991 to 2004, collected by the Station of Jianou Plant Protectin and Quarantine, Fujian Province, were used as study factors to build models. And then the data from 2005 to 2006 were used to test the accuracies of models. The results showed that prediction accuracies were over 99% by the optimized prediction models. These results demonstrate that this approach can effectively improve the prediction accuracy and does not depend on longtime series data.